Design and synthesis of benzopyran-based inhibitors of the hypoxia-inducible factor-1 pathway with improved water solubility

Abstract While progress has been made in treating cancer, cytotoxic chemotherapeutic agents are still the most widely used drugs and are associated with severe side-effects. Drugs that target unique molecular signalling pathways are needed for treating cancer with low or no intrinsic toxicity to normal cells. Our goal is to target hypoxic tumours and specifically the hypoxia inducible factor (HIF) pathway for the development of new cancer therapies. To this end, we have previously developed benzopyran-based HIF-1 inhibitors such as arylsulfonamide KCN1. However, KCN1 and its earlier analogs have poor water solubility, which hamper their applications. Herein, we describe a series of KCN1 analogs that incorporate a morpholine moiety at various positions. We found that replacing the benzopyran group of KCN1 with a phenyl group with a morpholinomethyl moiety at the para positions had minimal effect on potency and improved the water solubility of two new compounds by more than 10-fold compared to KCN1, the lead compound.


Introduction
Cancer is one of the leading causes of death, second only to heart disease 1 . One of the hallmarks of cancer is the formation of hypoxic areas inside of solid tumours 2 . This hypoxic tumour microenvironment leads to many changes such as the upregulation of pro-angiogenic and pro-glycolytic pathways, as well as increases in cell proliferation, genetic instability, and metastatic potential 3 . A major mediator of the hypoxic response is the hypoxia inducible factor (HIF) pathway 4 . HIF is a heterodimeric transcription factor consisting of two subunits, HIF-a, the stability of which is regulated by oxygen, and HIF-1b, which is constitutively expressed 5 . There are three known isoforms of HIF-a, HIF-1a, HIF-2a, and HIF-3a, with HIF-1a being the most commonly expressed and most extensively studied. Under normoxic conditions, HIF-a subunits are hydroxylated by a prolyl hydroxylase (PHD2) using molecular oxygen and then degraded via a VHL-dependent ubiquitination pathway 6 . Under hypoxic conditions, however, HIF-a subunits are stabilised, heterodimerise with HIF-1b and recruit coactivators such as p300 and CBP, to form active transcription complexes that bind to 5 0 -HREs (hypoxia response elements) in promoter regions of hypoxia-inducible genes 7 . Increased levels of HIF-1a are linked to cancer progression and poor patient outcome. Therefore, HIF is an attractive target for developing anticancer therapeutics 8 .
Further in vivo studies demonstrated 1's very pronounced inhibitory activity against brain, and pancreatic cancers 11 . In addition, 1 was well tolerated in mice; daily treatments with 60 mg/kg for up to 12 weeks had minimal side effects 11 . Neither did 1 nor its analogs demonstrate cytotoxicity, indicating the selective inhibitory effects being based on pathways unique to cancer 11 . Such results strongly suggest that this is a very promising class of compounds and warrant further studies. In fact, a previously synthesised and analysed class of analogs has been developed, which led to the discovery of 64b (Figure 1, 2, N-cyclobutyl-N-((2,2-dimethyl-2H-pyrano[3,2-b]pyridin-6-yl)methyl)-3,4-dimethoxybenzenesulfonamide) with an IC 50 value of $0.3 mM. 12 However, 1 and its analogs possess poor solubility in water (0.009 mg/mL) 11 . Therefore, dissolution in DMSO is necessary for in vitro assays and cremophor:ethanolbased formulations are needed for in vivo models. Such a formulation introduces undesirable properties 12 . It is well known that the successful development of potential therapeutics relies on more parameters than potency alone. Other properties, including solubility, can play a critical role. Therefore, we are interested in designing water-soluble analogs of 1 and 2 to address this critical aspect of drug development.

Synthesis
General methods and materials All commercial chemicals were of reagent grade from VWR (Radnor, PA), Aldrich (St. Louis, MO), or Oakwood Chemicals (Estill, SC), and were used without further purification unless otherwise indicated. 1 H and 13C spectra were obtained on a Bruker 400 NMR spectrometer at 400 and 100 MHz, respectively, in deuterated solvent with TMS (d ¼ 0.00 ppm) or deuterated solvent as internal reference. For all reactions, analytical grade solvent was used. Anhydrous solvents were used for all moisture-sensitive reactions. The Mass Spectrometry Facilities at Georgia State University obtained high-resolution mass spectra on a Waters Micromass Q-TOF (ESI) instrument.
Typical procedure for morpholine substitution (8a-c) Benzyl bromide (1 equivalent) was dissolved in acetonitrile. Morpholine (1.1 equivalents) and K 2 CO 3 (2 equivalents) were added and the reaction was stirred overnight at room temperature. The reaction was filtered through Celite and concentrated to give the product in quantitative yield. Typical procedure for reductive amination with aniline (10a-d, 14a) Aldehyde (1 equivalent), NaBH 4 (1.5 equivalents), and InCl 3 (0.15 equivalents) were dissolved in anhydrous ACN under inert gas. Aniline (1.5 equivalents) was added and the reaction was stirred until completion as monitored by TLC (typically $20 min). The reaction was quenched with saturated NH 4 Cl, taken up in ethyl acetate, washed with brine, dried over MgSO 4 , and concentrated. Column chromatography (1:1 hexane/ethyl acetate) was used to yield the final pure product.

Lipophilicity and solubility prediction
The in silico log D and log S values of all analogs were predicted using Calculator Plugins from MarvinSketch 4.3.0, 2017, ChemAxon (http://www.chemaxon.com), with results detailed in Table 1. Graphical representations of the log D and log S from pH 0 to 14 are provided in the Supplemental Information.

Luciferase assay
These analogs were first evaluated for their ability to inhibit hypoxia-induced HIF transcriptional activity in LN229-HRE-luciferase glioma cells as described previously 10-12 . Their inhibitory activities are presented as IC 50 in Table 1.

Solubility studies using dynamic light scattering
To further investigate the true enhancement of solubility, particle aggregation was examined using dynamic light scattering (DLS).
Selected compounds were treated according to the following procedure: 1. All centrifuge tubes and cuvettes were rinsed with either DCM or water and then vacuum dried before use to remove dust and any particulates.  Ã Experiments in PBS were carried out the same way as the experiments in water except DMSO stock solutions were made at 5 mM and the PBS diluted samples rested for 1 h before particle analysis.

Design
In considering ways to improve water solubility without compromising potency, we thought about introducing a commonly used morpholino moiety, which is known to help improve water solubility. In doing so, we were interested in searching for the optimal position, which would not negatively affect potency. Therefore, we devised four classes of compounds ( Figure 2): Class A incorporates a morpholinomethylphenyl or morpholinophenyl moiety instead of the 2,2-dimethyl-2H-chromene moiety and maintains the N-phenyl group; Class B incorporates either a morpholinomethylphenyl or morpholinophenyl moiety instead of the 2,2-dimethyl-2H-chromene moiety and substitutes the N-phenyl group for an N-cyclobutyl group; Class C has either a 2,2-dimethyl-2H-chromene or N-(2,2-dimethyl-2H-pyrano[3,2-b]pyridin-6-yl) moiety and either an N-ethylmorpholino or N-propylmorpholino group instead of the N-phenyl; and Class D has the 2,2-dimethyl-2H-chromene moiety with a N-phenyl-morpholine-4-sulfonamide.

Chemistry
Synthesis of Class A compounds (Scheme 1) was accomplished in four steps from 2-, 3-, or 4-bromomethylbenzylbromide 7a-c or in two steps from 4-morpholinobenzaldehyde 9d. Intermediates 7a-c were substituted with morpholine to yield morpholinomethylbenzylbromides 8a-c in quantitative yield. Next, the phenyl bromides 8a-c were converted to benzaldehydes 9a-c via lithium-halogen exchange at À78 C under inert gas. The aryllithium intermediate was treated with DMF as the electrophile in situ to generate the final benzaldehydes 9a-c. The aldehydes 9a-d underwent reductive amination with aniline to afford the secondary amines 10a-d.
Finally, 10a-d were reacted with 3,4-dimethoxybenzenesulfonyl chloride to afford sulfonamides 2a-d. Class B compounds (Scheme 1(C)) were synthesised in almost the same fashion as Class A, except that reductive amination of 9a-d was with cyclobutylamine and was not catalysed by any Lewis acid. Class C compounds were synthesised (Scheme 2) from 2,2-dimethyl-2H-chromene-6-carbaldehyde 12, which was readily synthesised from published procedures 13 . The aldehyde 12 underwent reductive amination with either ethylaminomorpholine or propylaminomorpholine to give secondary amines 13a-b, which were then reacted with 3,4-dimethoxybenzenesulfonyl chloride to afford sulfonamides 5a-b.
Class D compounds were synthesised (Scheme 3) from 12 in two steps. First, 12 underwent reductive amination with either aniline or cyclobutylamine to give secondary amines 14a-b. Next, the amines 14a-b were reacted with 4-morpholinosulfonyl chloride to afford sulfonamides 6a-b.

Biology
All the analogs were assessed for their ability to inhibit the HIF-1 pathway using a luciferase reporter assay described previously 10 . This assay reports the ability for a compound to inhibit HIF transcriptional activity. However, it does not specifically reveal the mode of action at the biochemical level. As can be seen from Table 1, introduction of a morpholino unit on the sulfonamide nitrogen led to compounds (5) with substantially diminished activity. The same is true if the morpholino unit is directly attached to the sulfonyl group (6). In the two series of compounds (3, 4) with a substituted phenyl group replacing the benzopyran ring in 1, only introduction of the morpholino moiety at the para positions (3) allowed for the preservation of HIF inhibition activity. Indeed, compounds 3a and 3d, which have exchanged the benzopyran ring for a para-morpholinomethylphenyl and para-morpholinophenyl, respectively, exhibit IC 50 values of 0.9 and 3.8 mM.
Similarly, analogs 4a and 4d, which replace the N-phenyl with a N-cyclobutyl, but are otherwise structurally the same as 3a and analogs synthesised in this work exhibited HIF inhibitory activity with IC 50 lower than 5 mM, suggesting the importance of conserving electronic and/or steric effects para to the phenyl ring. In particular, compounds 3a and 4a are active within the same order of magnitude as 1, and are about threefold less active than the previously discovered 2 (IC 50 ¼ $0.3 mM) 12 . The improved potency of 3a and 4a over 3d and 4d suggests a possible role for flexibility of the ligand in the binding site.
To gain some initial understanding of lipophilicity and solubility, the predicted log D and log S values were calculated for 1, 2, and their analogs. Log P refers to a molecule's partition coefficient, or the log of the ratio between its solubility in octanol versus water 14 . This is commonly used to indicate a candidate drug's lipophilicity, and a log P or cLog P (calculated log P) less than 5 is generally considered "drug-like" 15 . For ionizable small molecules, log D is the distribution constant, which describes the partition coefficient at different pH levels 16 . A molecule's water solubility is typically measured at room temperature (20-25 C) in mol/L and represented as log S, or clog S when calculated computationally. Drugs on the market with a variety of structures typically possess a log S between À5 and À2 17 .
Though several of the morpholine analogs have very drug-like properties, most are not active in the luciferase assay. Only 3a, 3d, 4a, and 4d are active toward the HIF pathway and only 3a and 4a show comparable IC 50 values as 1. Therefore, we examined their solubility in water and phosphate buffered saline (PBS).

Solubility studies
To investigate the true enhancement of aqueous solubility, particle aggregation was examined using DLS. DLS can detect particle sizes in solution by measuring changes in scattered light in relation to the Brownian motion of particles 18 . It is commonly used to detect the particle sizes of various chemical and biological molecules, including small molecule inhibitors 19 . Though there are several methods for detecting solubility, we chose the DLS method due to its ease, reproducibility, minimal sample requirement, and relative sensitivity to small particles.
The active compounds 3a and 4a were compared to their nonmorpholine containing counterparts, 1 and 2, respectively ( Figure 3). Solutions of varying concentrations of each compound were made in either water or PBS with 1% DMSO. Each solution was measured in the particle size analyser to identify which samples showed formation of aggregates in solution. DLS measurements, summarised in Table 2, reveal that 3a forms aggregates at approximately 10 mM, an order of magnitude higher than 1, which is insoluble at just 1 mM in water. The N-cyclobutyl analog 4a forms aggregates in excess of 100 mM, significantly higher than its counterpart 2, which forms particles at a mere 10 mM. In PBS, the solubilities parallel those seen in the water solution, where 1 and 3a exhibit comparable particle formation at 1 mM and 15 mM, respectively. 2 shows particle formation at 10 mM, while 4a shows none at this concentration, as expected. Indeed, with a log D of 2.94 log S of À3.36 (Table 1), 4a is predicted to be quite soluble in aqueous solutions.
The described results clearly indicate that (1) the benzopyran ring can be modified with minimal loss of activity and (2) the para position of the phenyl ring can tolerate substantial changes and can be used for improvement of water solubility. Such results will help future optimisation work.

Conclusion
Of the 12 new morpholine-containing analogs developed in this work, four demonstrate HIF inhibition in the low or sub-micromolar range. In particular, 3a and 4a both exhibit inhibition of HIF transcriptional activity with IC 50 values of 0.9 and 1.0 mM, respectively. As expected, the in silico log P and log S values of these analogs are considered more favourable than lead compound 1 or its more potent analog 2, and are therefore likely to be more bioavailable. Following these indications, solubility as measured by particle detection with DLS reveal the exceptional solubility of analogs 3a and 4a over their non-morpholine containing predecessors 1 and 2. Particle formation of 4a is undetected in excess of 100 mM in water and 50 mM in PBS, while still displaying HIF inhibition in the same order of magnitude as lead 1. These results encourage exploration and use of more soluble moieties to further probe the SAR (structure-activity relationship) and SSR (structure-solubility relationship) of potential analogs.